Real Time Implementation of Cartesian Space Control

for Underactuated Robot Fingers

Jamaludin Jalani

Department of Electrical Engineering Technology, Faculty of Engineering Technology, Universiti Tun Hussein Onn
Malaysia, 86400, Parit Raja, Batu Pahat, J ohor, Malaysia.
jamalj@uthm.edu.my


Abstract This paper presents a real time implementation of a sliding mode controller
(SMC) for the underactuated BERUL
1
fingers. The use of SMC is to deal with the
inaccuracies and unmodelled nonlinearities in the dynamic model of the robotic fingers, in
particular, to overcome significant friction and stiction. This paper in particular shows a
practical implementation of Cartesian space control for the BERUL fingers. For the
purpose of comparison, the performance of the PID controller is included in this paper.
The main controller for positioning control is the combination of a feedback linearization
(FL) scheme and SMC. The results show that the SMC performed better than the PID
controller in the Cartesian space control.

Index Terms Robot Fingers, Sliding Mode Controller, PID Controller, Cartesian Space
Control
I. INTRODUCTION
Various techniques have been proposed to grasp an object for robot fingers. According to Montana [1], we
can divide the general approaches to robot fingers manipulation into two categories, those which focus solely
on velocity and those which deal with configuration-space issue. He further describes the velocity based
approaches into four categories namely purely kinematic control, dynamic control, control the location of the
object and control the position of the point of contact. A purely kinematic control means that the state of
the finger plus object system is completely determined by the state of the finger joints. In contrast to the
kinematic control, the dynamic control means that the state of the full systemdepends on dynamics quantities
as the state of the finger joints. The location of the object reflects to the center of its own body whilst the
position of the point of contact implies the contact surface of its own body. In general, previous work on
velocity-based control can be found in Montana [2, 3, 4, 5] and Hunt et al. [6]. On the other hand, we can
define the configuration space as the space of possible positions that a robot finger system may attain. Thus,
this requires the set of all legal combinations of finger joint angles and contact states of an object. Some
work on the configuration space issue can be referred to are in Fearing [7], Hong et al. [8], Li and Canny [9]
and Brock [10].
Hence, to investigate a grasping performance for the BERUL fingers, a velocity based approaches as
mentioned above is considered. The robust control technique i.e. a sliding mode controller (SMC) is used to
overcome friction and stiction. It is to note that the BERUL fingers are significantly affected by stiction. The
work in [11, 12, 13, 14] has been suggested to overcome friction and stiction where an adaptive controller was
1
The mechanical design and manufacturing for the BERUL hand has been conducted by Elumotion (www.elumotion.com).

DOI: 01.IJRTET.10.2.1377
Association of Computer Electronics and Electrical Engineers, 2014

Int. J. on Recent Trends in Engineering and Technology, Vol. 10, No. 2, Jan 2014

158
employed. Moreover, a new control approaches via a Nominal Characteristic Trajectory Following (NCTF)
[15] was proposed mainly to overcome friction and stiction for the BERUL fingers. The results have shown
that the NCTF produced lower total signal energy as compared to the adaptive control. However, the
controller has not been tested experimentally for the robot fingers. It is important to show that the grasping
can be achieved in real time. In thi s work, a wel l known sl i di ng mode control l er (SMC) i s
proposed and i nvesti gated. The implementation of the SMC is carried out based on the forward
kinematics as provided in [16]. Having found the forward kinematics of the BERUL fingers, the Cartesian
space control can be introduced.
The kinematics of the manipulators contains position, orientation, and velocity analysis of manipulators which
can be exploited for tracking and grasping of the BERUL fingers. For this, the J acobian of the forward
kinematics has been deployed in order to implement a Cartesian coordinate space control. Specifically, the
Jacobian can be described as the configuration of the finger changes while in motion, the mapping of
velocities changes accordingly. Having the kinematics of the BERUL fingers, the Cartesian coordinate
space control can be achieved by employing Sliding Mode Control (SMC) [17]. At this point, the Jacobian
is also used for mapping of forces and torques where the set of joint torques are required to generate the
force and moment at the end-effector. The success in the Cartesian coordinates control will lead to the
implementation of a compliance control in a cylindrical coordinate systemin future. In general, this paper is
organized as follows. Section II focuses on the summary of the kinematics of the BERUL fingers. Section
III introduces controller design for which the Sliding Mode Controller (SMC) is employed to alleviate
nonlinearities. Section IV illustrates the experimental set up, Section V discusses the tracking performance
and followed by conclusions in Section VI.
II. FORWARD KINEMATICS ANALYSIS
It is very important to show that a sufficient forward kinematics of the BERUL hand is obtained. A simple
way of modeling kinematics is by using the Denavit Hartenberg (D-H) representation. The D-H technique
allows us to use a systematic approach for deriving the forward kinematics in particular for complex
transformations which consist of several joints and links. In brief, the following forward kinematics is
obtained (see Figure 1) for the end-effector (i.e. [X, Y, Z]):

X =o
1
cos0
1
+o
2
cos (0
1
+0
2
) (1)

=o
1
sin0
1
+o
2
sin (0
1
+0
2
) (2)

Z = 0 (3)
1

2
a
1
a
1
L
in
k
2
L
i
n
k
) , , ( Z Y X
0
Y
0
Z
0
X
1 Y
1 Z
1 X
2
Y
2
Z
2
X

Figure 1. Derivation of forward kinematics for two links robot manipulator by using DH technique
Moreover, we can also derive the kinematics Jacobian fromequations (1) and (2) as follows:

159

oX
o0
1
=o
1
sin0
1
o
2
sin(0
1
+0
2
)
(4)

o
o0
1
=o
1
cos0
1
+o
2
cos (0
1
+0
2
)
(5)

oX
o0
2
=o
2
sin (0
1
+0
2
)
(6)

o
o0
2
=o
2
cos (0
1
+0
2
)
(7)

z
0
1
= 0
(8)
oZ
o0
2
= 0
(9)
Similarly, the forward kinematics of the BERUL fingers can be easily derived by using the D-H technique. A
sufficient of the forward kinematics of the BERUL finger has been shown in [16]. Apart fromusing D-H
technique, Jalani [16] also introduced a new technique via Roborealm to obtain a sufficient forward
kinematics of the BERUL fingers. The forward kinematics results obtained in [16] can be practically used for
positioning control in the Cartesian space. Subsequently, the Jacobian can be easily obtained once the forward
kinematics is available.
III. CONTROLLER DESIGN
The structure of the proposed controller for the BERUL fingers is shown in Figure 2. The function of the
controller is to control the finger position for example X coordinate so that it moves to the desired position
X
d
.
d
X
q q,

+
-

X
F

Figure 2. The structure of the proposed controller
For controller design, the model

q + =u (10)
is considered, where m is the generalized mass/inertia, is a lumped expression for the major nonlinearities
i.e. gravity, friction and centrifugal/coriolis force.
A. Sliding Mode Controller (SMC)
Motivated by the recent development of the SMC on a two-link rigid robotic manipulator through simulation
[18], the same controller is applied for the BERUL fingers. Using the general dynamic equation of (10), a
suitable sliding mode controller [18] is designed as follows. The joint torque vector can be split into two
additive terms:
where m
0
(q),
0
(q,q)

are the nominal value of m(q) and (q,q)

respectively; which are defined as

m
0
(q) =m(q) m and
0
(q) =(q) ; K
P
R
nxn
and K

R
nxn
are positive scalars
determining the closed loop performance; and the tracking error is defined as q
c
(
t
)
=q(
t
)

q
d
(
t
) with
[q
d
(
t
)

q
d
(
t
)

q
d
(
t
)
]
being the reference trajectory and its time derivatives. Note that
0 represents the
computed torque component and
1
denotes a discontinuous torque control. The discontinuous torque
control and sliding manifold are respectively defined as
=
0
+
1
(11)

160

1
=
0
sign(s)

(12)
s =q
c +
K
s
q
c


(13)
In order to control the effect of the chattering, the following termis used,
s
s +|o|

(14)
instead of sign(s) where o allows to suppress chattering.
B. PID Controller
For the purpose of comparison, a PID control scheme is adopted from Ge et al. [14] and Jalani et al. [15].
Then, the control input term(fromequation (14) as given in [15]) can be simplified as follows.
u =k
1
+k

_rJt
t
0


(15)
where k
1
represents a proportional gain and k

represents an integral gain. Note that, using r fromequation

(11) as given in [15], the control in (8) is indeed a PID controller.
C. Cartesian Space Control
The derivation of the robot finger kinematics as elaborated in Section 2 allows us to express the finger
position in the Cartesian coordinates. For example, the position x =x(q) of the thumb finger can be
uniquely computed as a function of q where the relationship can be expressed in Cartesian velocities and
accelerations respectively as

x =[q (16)


[ is the Jacobian of the kinematics x =x(q) i.e.

0
= m
0
(q)(q
d
K

q
c
K
P
q
c +

0
(q,q )),
(18)
The obtained Jacobian will be used for the following applied torque
u =[
1
F

(19)
where F is the actuator of the finger. This implies that the dynamics of (10) can be rewritten as:
[
-1
m[
-1
x [
-1
m[
-1
[

q +[
-1
=[
-1
u

(20)
or
mx +
`
=F

(21)
where
m =[
-1
m[
-1
(22)
and

`
=[
-1
m[
-1
[

q +[
-1
(23)

and
F =[
-1
u

(24)
This new model allows the sliding mode approaches of Shi et al. [18] to be used.
IV. EXPERIMENTAL SETUP
For the real time implementation of the control scheme, a dSPACE DS1106 embedded systemis employed.
The Simulink model of the scheme is compiled into real-time C code and run in the dSPACE system. A
sampling time of 1 millisecond is used. The BERUL fingers use optical encoders for position and velocity.
x =[

q +[q (17)

161
Each joint is equipped with a Maxon EPOS programmable digital motor controller. All the sensors and
actuators of each joint are connected to the EPOS positioning controllers, which communicate with dSPACE
via a CAN-Bus and provide the low-level current control for torque demands with the brush-less DC motors.
A multi-step demand is applied in y and z direction. In general, Figure 3 shows the experimental setup for
the BERUL fingers.

Figure 3. Experimental setup for theBERUL fingers
V. TRACKING PERFORMANCE
Based on a given workspace as shown in Figure 4, the tracking in Cartesian space is practically illustrated in
y direction for index, middle, ring and small fingers. In contrary, the thumb is practically depicted in either x
and/or z direction.

Figure 4. Workspace for the BERUL fingers
Here, only ring and thumb fingers are practically shown in Cartesian space control. Generally, the results
show that the PID and the SMC follow a required trajectory satisfactorily as shown in Figure 5(a), Figure
5(b), Figure 6(a) and Figure 6(b) respectively. The detailed performance fromeach controller can be seen
by using a quantitative measure of performance as indicated by [19] and [20] which is computed as follows:
NormolizcJ Error =|yJi yi|/ N
s
N
s
=0


(25)
while we also consider
NormolizcJ Error =|u

|
2
N
s
=0


(26)
where yJi is the demand signals, yi is the output signal, N
s
is the number of samples and u

is the control
signal.
-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
-0.1
-0.05
0
0.05
0.1
0.15
-0.1
-0.05
0
0.05
0.1
0.15
Y axis
X axis
Z

a
x
i
s
index
workspace
ring
workspace
small workspace
thumb
workspace
middle
workspace

162

(a)Ring finger for y position

(b)Thumb finger for z position
Figure 5. Cartesian Space Control for PID

(a)Ring finger for y position
0 10 20 30 40 50 60 70 80 90 100
-0.2
-0.1
0
Time (sec)
(
m
)


Demand
Actual
0 10 20 30 40 50 60 70 80 90 100
-0.02
0
0.02
Time (sec)
E
r
r
o
r
(
m
)
0 10 20 30 40 50 60 70 80 90 100
-200
0
200
Time (sec)
C
o
n
t
r
o
l

S
i
g
n
a
l

(
m
A
)
0 10 20 30 40 50 60 70 80 90 100
-0.04
-0.02
0
Time (sec)
(
m
)


Demand
Actual
0 10 20 30 40 50 60 70 80 90 100
-5
0
5
x 10
-3
Time (sec)
E
r
r
o
r
(
m
)
0 10 20 30 40 50 60 70 80 90 100
-1000
0
1000
Time (sec)
C
o
n
t
r
o
l

S
i
g
n
a
l

(
m
A
)
0 10 20 30 40 50 60 70 80 90 100
-0.2
-0.1
0
Time (sec)
(
m
)


Demand
Actual
0 10 20 30 40 50 60 70 80 90 100
-0.02
0
0.02
Time (sec)
E
r
r
o
r
(
m
)
0 10 20 30 40 50 60 70 80 90 100
-200
0
200
Time (sec)
C
o
n
t
r
o
l

S
i
g
n
a
l

(
m
A
)

163

(b)Thumb finger for z position
Figure 6. Cartesian Space Control for SMC
TABLE I. NORMALI ZED ERROR AND TOTAL CONTROL ENERGY FOR THUMB AND RING FINGERS
Finger Controllers Normalized
Error (rad)
Total Control
Energy (mA)

Thumb

PID

0.0022

2.2184 10
5


Thumb

SMC

3.7260 10
-4


3.6717 10
5


Ring

PID

0.0054

1.9231 10
3


Ring

SMC

0.0026

1.3934 10
3

It is found that the SMC has the smallest normalized error for the thumb and the ring finger. However, the
PID controller uses smaller total control energy for the thumb finger and the SMC controller produces
smaller total control energy for the ring finger. The details of the error signal and total energy produced by
each controller can be referred to in Table 1.
VI. CONCLUSIONS
In this paper, the kinematics of the constrained underactuated BERUL fingers is used for Cartesian space
control. It has been shown that the Cartesian space control was successfully demonstrated in real time
implementation. In addition, the SMC is important to overcome the nonlinearities and uncertainties in
particular friction and stiction which are emanated fromthe BERUL fingers. It is found that the SMC
provides better normalized error for the ring and the thumb finger. However, the total control signal energy
is slightly higher for the thumb finger.
ACKNOWLEDGMENT
The CHRIS (Cooperative Human Robot Interaction Systems) project is funded by the European
Commissions Seventh Framework Programme (FP7) and run from2008- 2012. The research is also partially
funded by the Malaysian Government.
REFERENCES
[1] D.J . Montana. The kinematics of multi-fingered manipulation. In IEEE Transactions on Robotics and Automation,
volume 11, pages 491503, aug 1995.
[2] David J . Montana. Contact stability for two-fingered grasps. In IEEE Transactions on Robotics and Automation,
0 10 20 30 40 50 60 70 80 90 100
-0.04
-0.02
0
Time (sec)
(
m
)


Demand
Actual
0 10 20 30 40 50 60 70 80 90 100
-2
0
2
x 10
-3
Time (sec)
E
r
r
o
r
(
m
)
0 10 20 30 40 50 60 70 80 90 100
-2000
0
2000
Time (sec)
C
o
n
t
r
o
l

S
i
g
n
a
l

(
m
A
)

164
volume 8, pages 421 430, aug 1992.
[3] David J . Montana. The kinematics of contact and grasp. The International J ournal of Robotics Research, 7(3):17
32, 1988.
[4] D.J . Montana. Tactile sensing and the kinematics of contacts. PhD thesis, 1986.
[5] D.J . Montana. The kinematics of contact and grasp. International J ournal Robotics Research, 8(4):421430, aug
1995.
[6] K.H. Hunt, A.E. Samuel, and P.R. McAree. Special Configurations of Multi-finger Multi- freedomGrippers- A
Kinematic Study. The International J ournal of Robotics Research, 10(2):123134, 1991.
[7] R. Fearing. Implementing a force strategy for object re-orientation. In Proceedings of the 1986 IEEE International
Conference on Robotics and Automation, volume 3, pages 96 102, April 1986.
[8] J . Hong, G. Lafferriere, B. Mishra, and X. Tan. Fine manipulation with multifinger hands. In IEEE International
Conference on Robotics and Automation, volume 3, pages 1568 1573, May 1990.
[9] Z. Li and J . Canny. Motion of two rigid bodies with rolling constraint. volume 6, pages 62 72, February 1990.
[10] D. L. Brock. Enhancing the dexterity of a robot hand using controlled slip. In IEEE International Conference on
Robotics and Automation, volume 1, pages 249 251, April 1988.
[11] C. C. de Wit, K. Astrom, and K. Braun, Adaptive friction compensation in dc-motor drives, IEEE J ournal of
Robotics and Automation, vol. 3, no. 6, pp. 681685, December 1987.
[12] C. C. de Wit, H. Olsson, K. Astrom, and P. Lischinsky, Dynamic friction models and control design, American
Control Conference, pp. 19201926, J une 1993.
[13] C. C. de Wit and S. Ge, Adaptive friction compensation for systems with generalized velocity/position friction
dependency, Proceeding of the 36th IEEE Conference on Decision and Control, vol. 3, pp. 24652470, Dec 1997.
[14] S. Ge, T. Lee, and S. Ren, Adaptive friction compensation of servo mechanisms, International J ournal of Systems
Science, vol. 32, no. 4, pp. 523532, 2001.
[15] J alani, J ., NCTF Control for Friction Compensation of Underactuated Fingers, International. J ournal on Control
System and Instrumentation, ACEEE, Vol. 03, No. 01, Feb 2012.
[16] J alani, J ., Development of the Forward Kinematics for Robot Fingers By Using Roborealm, International J ournal
of Research and Reviews in Applied Sciences, May 2011, Volume 7, Issue 2.
[17] V. I. Utkin. Sliding Modes in Optimization and Control,. Springer-Verlag, New York, 1992.
[18] J . Shi, H. Liu, and N. Bajcinca. Robust control of robotic manipulators based on integral sliding mode. International
J ournal of Control, 81:15371548 vol.81, 2008.
[19] D.J .G. J ames S.G. Goodhart, K.J . Burnham. Self-tuning control of non-linear plant - a bilinear approach. In
Transactions of the Institute of Measurement and Control, volume 14, pages 227232, 1992.
[20] C. Edwards and S.K. Spurgeon. Robust nonlinear control of heating plant. In IEE Proceedings on Control Theory
and Applications, volume 141, pages 227234, J ul 1994.